• AI visibility

Guide · AI search

Do Google AI Overviews reduce clicks to my website?

Yes, on the best evidence available. In a 2026 field experiment with 1,100 US Google users, hiding AI Overviews raised the rate of clicks to outside websites by 8.8 percentage points. Browsing data from Pew researchers and a study of Wikipedia traffic point the same way, though the size of the loss depends on the site and the searches.

By Underneath Team Guide Updated

The short version

  1. Hiding AI Overviews raised clicks to outside websites by 8.8 percentage points per search in a preregistered March 2026 experiment with 1,100 US users (Wang and colleagues).
  2. In a Pew panel of 900 US adults in March 2025, people clicked a result on 8% of visits to pages with an AI Overview, against 15% on pages without one.
  3. People in the same panel ended their browsing session on 26% of pages with an AI Overview, against 16% without.
  4. English Wikipedia lost about 5% of its search traffic once AI Overviews became Google’s US default, roughly 100.27 million visits a month (Khosravi and Yoganarasimhan).
  5. The clicks lost bought no measurable gain for users: in the 1,100-person experiment, hiding AI Overviews changed neither trust in Google nor satisfaction with search.

What is the strongest evidence that AI Overviews cut clicks?

A 2026 randomized field experiment on everyday Google use found that hiding AI Overviews raised outside clicks.

Wang, Gleason, Bart, Wilson and Metaxa (opens in a new tab) recruited US adults who search with Google in Chrome and installed a browser extension. After three normal days, people were randomly assigned to one of three versions of Google for seven days: AI Overviews hidden, Google unchanged, or every search sent to AI Mode. Before the switch, 36% of their searches showed an AI Overview.

Hiding AI Overviews raised the click-through rate, meaning clicks to outside websites divided by searches, by 8.8 percentage points. The authors also saw suggestive evidence of extra clicks to news publishers. One complication: a Google page change broke the hiding tool partway through, so only 51.1% of AI Overviews were actually hidden. The 8.8-point estimate is adjusted for that.

The sample was not typical of the US: 75% were under 45 and 87% had at least some college education. The test lasted a week, so it says nothing about how habits settle over months.

What does real-world browsing data show?

Pew’s browsing panel shows people clicked results about half as often when an AI Overview appeared.

Chapekis, Lieb, Shah and Smith (opens in a new tab) at Pew Research Center tracked the web browsing of 900 US adults in March 2025, covering 68,879 distinct Google searches. This is observational data, so it shows an association, not proof of cause. The gaps held after the authors allowed for the type of search, such as length and question words.

What people did nextPage with an AI OverviewPage without one
Clicked a search result8%15%
Ended their browsing session26%16%
Clicked a source cited in the AI Overview1%not applicable

A separate industry analysis, reported by Aral, Li and Zuo (opens in a new tab), found a median zero-click rate of 80% on searches with an AI Overview against 60% without. We could not review that analysis directly. We look more closely at why people stop browsing after an AI Overview in a separate guide.

How much traffic did a real publisher lose?

English Wikipedia lost roughly 5% of its search traffic after AI Overviews became Google’s US default.

Khosravi and Yoganarasimhan (opens in a new tab) compared search visits to about half a million English Wikipedia articles with the same articles in German and French. Google made AI Overviews part of normal US search in May 2024, while they were not part of default search in the German and French markets during the study period. The study covers December 2023 to December 2024.

English search traffic fell by about 5.45% relative to German and 4.82% relative to French. That works out to roughly 100.27 million fewer search visits a month. A comparison with Japanese Wikipedia, which the authors treat as a weaker check, gave a 16.53% decline.

These are averages across all searches, all search engines and all users, including the many who never saw an AI Overview. The authors note that the loss on searches that actually showed one would be larger. Several other papers quote a figure of about 15% for this study; the sixth version, which we reviewed, reports 5.45% and 4.82%. The Wikipedia case shows how little control publishers have when Google changes its results pages.

Is the loss the same for every site?

No: it depends on the site, the searches it relies on and whether its content can be summarized.

Wikipedia is a hard case for AI Overviews to hurt, because it is among the sources they cite most. It still lost traffic. Whether being cited makes up for lost clicks is a question we cover separately. Khosravi and Yoganarasimhan note that a separate experiment by Agarwal and Sen found larger effects, possibly because ordinary publishers lose more than Wikipedia. Xu, Iqbal and Montgomery (opens in a new tab) summarize that experiment as cutting organic click-throughs by 38%; we could not review it directly.

Not every site loses. Zhang, Cui and Zhang (opens in a new tab) found that Reddit communities AI Overviews can cite gained 12.0% more daily comments than communities Google excludes from them. The gains were largest for opinion and experience discussions, which a summary cannot replace.

The stakes are commercial. Xu and colleagues found that 50.63% of the pages AI Overviews cited carried display ads, so a lost click is often lost revenue as well as lost attention.

Do users gain anything in exchange for fewer clicks?

Not measurably: hiding AI Overviews changed neither trust in Google nor satisfaction in the experiment.

Wang and colleagues found no detectable change in trust, usefulness, satisfaction or sense of control when AI Overviews were hidden. Besides clicking, the clearest change was that people spent 0.59 fewer minutes per search session when AI Overviews were hidden.

That does not prove AI Overviews are useless to searchers. Khosravi and Yoganarasimhan stress that their data measures publisher traffic, not the value of users’ time. Some people do get their answer faster. But in the only randomized test of real use, the clicks publishers lost were not matched by a gain users noticed.

Is AI Mode worse for clicks than AI Overviews?

Yes: sending every search to AI Mode cut outside clicks by 18.8 percentage points, about twice the AI Overview effect.

AI Mode is Google’s conversational search, a separate tab where Gemini answers and takes follow-up questions. In the same experiment, forced AI Mode also cut the share of people clicking through to news sites, Reddit and Wikipedia. Our article on what AI Mode as Google’s default would cost covers that result in detail.

What should you do about it?

Plan for fewer clicks per search on informational queries, and measure where your own exposure sits.

  1. Find which of your traffic-driving searches show AI Overviews. Our frequency study found them on 60.8% of 800 US commercial keywords, but far fewer on local searches.
  2. Forecast with a range, not one number. The evidence runs from about 5% of search visits for Wikipedia, averaged across all searches, to 8.8 points of click-through per search in the experiment.
  3. Track click-through, not just rankings. A page can hold position one and still lose clicks when a summary sits above it.
  4. Watch AI Mode. Its effect on clicks was about twice that of AI Overviews.
  5. Look at what your content offers beyond facts. Opinion and experience content gained engagement on Reddit, while factual reference content lost visits. For publishers, the risk to content businesses is covered in depth.

If you want help sizing your exposure, see our generative engine optimization service.

What does the research not tell us yet?

The research shows clicks fall, but not yet how much a typical business site loses in revenue.

  • Only one randomized experiment on real Google use was available to us, with a young, educated sample over seven days.
  • The Pew panel is observational and from March 2025; AI Overviews have changed since.
  • The Wikipedia study covers one nonprofit publisher in 2024.
  • No study here measures leads, sales or revenue for ordinary business sites.
  • Long-run effects are unknown: people may click more, or less, as habits settle.

Frequently asked questions

Do AI Overviews reduce organic traffic?

Yes, on average. Hiding them raised clicks to outside sites by 8.8 percentage points in a 2026 US experiment, and English Wikipedia lost about 5% of its search traffic after they launched.

How much do AI Overviews reduce click-through rate?

Pew’s 2025 browsing data shows clicks on results fell from 15% of visits to 8% when an AI Overview appeared. That is close to half, though the data shows an association rather than proof.

Do people click the links inside AI Overviews?

Rarely. In Pew’s panel, just 1% of visits to pages with an AI Overview led to a click on a cited source.

Do AI Overviews hurt every website equally?

No. Reddit discussion communities gained 12.0% more daily comments after AI Overviews launched, while factual sources such as Wikipedia lost visits.

Sources

Free strategy call

Some questions are easier to answer about your own business.

Bring the one that matters most. On a free 30-minute call we’ll take a first look at it and send you a short written read afterward.

Ask about my businessHow we work